Skene
Skene is a product intelligence agent that audits a software product's analytics tracking, checks every release for broken or missing events, and recommends journey improvements. It protects the collection layer feeding PostHog, Mixpanel and similar tools.
What is Skene?
Skene is a product intelligence agent that sits between the go-to-market teams who need product data and the engineering team that ships the product. The problem it addresses is narrow and well defined: a release renames an event, drops a field or changes when tracking fires, and nothing turns red. Dashboards keep updating, marketing sees a conversion dip, engineering sees a green release, and nobody can say whether customer behaviour changed or the collection did. By the time the gap is found, the missing history cannot be recreated.
The product runs three jobs in a loop. First it audits: it scans the product's collection layer, maps the signals already collected, flags missing steps, broken payloads and silent gaps, and prioritises fixes by the journeys they affect. Then it protects: every release is checked against the agreed collection plan, catching five distinct failure modes, since only a removed call leaves a visible hole while renamed, moved, altered and conditionally fired calls all still run. Finally it improves, recommending the next signals and automation opportunities, and running an Evaluator that says whether a measurement plan is ready to launch before it launches, then re-runs its success formula against real data afterwards.
One engine is exposed through four surfaces: a hosted MCP server a coding agent calls before opening a pull request, a GitHub App that reviews every pull request and carries a CI action that can fail the build, a cloud API that keeps a record between runs, and an MIT-licensed command-line tool that needs no Skene account. Skene Cloud adds an editable journey canvas, an events table, Flows showing how users move, and the Evaluator.
Crucially, Skene is not an analytics product. It does not store, model or visualise event data: it protects the tracking calls feeding PostHog, Mixpanel and whatever else a team already runs. It connects to GitHub for one repository and branch, and to Supabase over OAuth, read-only by default, reading table, column and type definitions rather than rows.
What it does
- Inventory the tracking signals a product already collects, and flag the broken, incomplete and missing ones
- Check every release against the collection plan and catch removed, renamed, moved, altered or conditionally fired events
- Comment on the exact line of a pull request and attach a one-click GitHub suggestion or a copy-pasteable fix prompt
- Fail a CI build when an event the plan requires is missing from the code, or the table behind one is gone
- Answer a coding agent over a hosted MCP server before it opens the pull request
- Say whether a measurement plan is ready to launch, and re-run its success formula against real data afterwards
- Recommend the next signals and automation opportunities for growth and customer success teams
When to use Skene / When not to
A quick filter to help you decide if Skene is the right fit.
When to use Skene
- Growth, revenue operations and customer success leaders who have to defend an activation or funnel number they did not collect themselves
- Product-led SaaS teams whose tracking quietly breaks on every release without anything turning red
- Engineering teams shipping with coding agents, which drop the tracking call as a side effect of a refactor
- Teams already running PostHog or Mixpanel who want the events feeding them to stay complete and correctly named
- Startups on a GitHub and Supabase or PostgreSQL stack looking for a self-serve check rather than an analytics consultancy
When not to use Skene
- Teams hosting their code on GitLab or Bitbucket, since the GitHub App is the only supported repository host
- Anyone looking for an analytics product: Skene does not store, model or visualise event data and will not replace PostHog or Mixpanel
- Organisations whose procurement requires a SOC 2 certificate or a published Data Processing Agreement today
- Teams whose analytics stack is neither PostHog nor Mixpanel, as those are the only two providers read as sources
- Anyone expecting a mobile app or a no-code product, since the tool lives in a browser dashboard, a terminal and a pull request
How to use Skene
A typical end-to-end flow, from setup to results.
- Sign in with a magic link; Skene creates your workspace on first login, with no password to set
- Install the Skene GitHub App, pick which repositories it can see, and confirm the one you want
- Or connect a Supabase project over OAuth instead, read-only by default; one source is enough to continue
- Run the first analysis to map what you collect today and see the broken, incomplete and missing signals
- Agree the customer journey with your go-to-market team so Skene can derive the collection plan
- Turn on pull request reviews from the switch under Settings, GitHub, to check every release from then on
- Read the findings on the pull request, apply the one-click GitHub suggestion, or paste the supplied prompt into your coding agent
- Comment /skene fix on a pull request to restore tracking calls that the pull request removed
- Add the hosted MCP server to your agent configuration, pointing at the endpoint with a workspace API key header
- Or work entirely locally instead with the open source CLI, running the analysis against a model provider you supply
Pros & Cons
Pros
- The CI gate is a deterministic comparison rather than a model call, so it keeps working even once the monthly allowance is used up
- Every finding is reviewable: the tool names the signal, the file and the reason it failed, instead of returning a score
- Nothing is pushed directly to your default branch and nothing enters your dependencies or lockfile
- The command-line half is open source under the MIT licence and runs with no Skene account and no service behind it
- Strong reversibility: uninstalling the GitHub App revokes all four permissions at the source, and the journey file already committed stays in your repository
- The Supabase connection starts read-only, reads no row data, never writes, and blocks data previews outright
- Ten named subprocessors are published with what each one sees, and the first audit is free with no card and does not expire
Cons
- GitHub is the only supported repository host: GitLab and Bitbucket are explicitly out
- Exactly two analytics providers can be read as sources, PostHog and Mixpanel, and nothing else
- No SOC 2 certification: the vendor states it is on the roadmap, not yet started, with no date to give
- No Data Processing Agreement is published or offered, and the personal data recipient list is only available on request
- The privacy policy declares training and optimising AI models on personal data under legitimate interest, with no documented way to opt out
- The pull request review is a model reading a diff: the vendor admits it can be wrong and can disagree with the deterministic gate
- The free tier is one audit per workspace only, Enterprise carries no public price, and the vendor is a five-person company incorporated in August 2025
Pricing & Plans
A permanent free plan is available: it costs USD 0, requires no payment card, and provides one product data collection analysis per workspace, which does not expire. The lowest paid price point is the Pro plan at USD 249 per month. Paying annually reduces this to USD 2,390.40 per year, a twenty percent saving against USD 2,988 paid monthly. The Enterprise tier is quoted rather than listed and carries no public price. The open source command-line tool is free of charge and requires no account, provided the user supplies their own model provider.
- one audit per workspace
- once
- with no card
- shows the broken
- incomplete and missing signals in the current product journey
- includes a monthly usage allowance
- checks on every pull request
- the next collection and journey improvements as they appear
- and the workspace owner emailed at 80% and 100% of the allowance
- expert-curated rules and guidance
- a shared go-to-market and engineering operating model
- and adaptation to a complex stack
- Skene OSS — free
- outside the tiers — the MIT-licensed CLI and terminal UI
- with no Skene account and a model provider you supply
Data, GDPR & hosting
A consolidated view of how Skene handles your data.
GDPR overview
GDPR compliance is claimed explicitly: the privacy policy states the company complies with the GDPR in all processing of personal data, alongside applicable national legislation. Concrete mentions are present. Skene Technologies Oy is named as controller, with DPO@skene.ai as the contact for data subject requests, which must be made in writing and refused only with written reasons. Legal bases are listed purpose by purpose: contract, legitimate interest, consent and statutory obligation. Rights covered include access, rectification, erasure, portability, restriction, objection and withdrawal of consent. Servers and data are said to be mainly hosted within the European Union, with exceptional transfers relying on adequacy decisions or standard contractual clauses. No automated decision-making is carried out. The competent supervisory authority named is the Finnish Data Protection Ombudsman. The policy was published on 30 October 2025, version 1.0.
Who owns the data?
Under the Terms of Service, the customer keeps all intellectual property rights to the User Content submitted, meaning the code, configurations and documents Skene analyses. Skene receives only a non-exclusive licence, limited to running the service, producing the Results, providing support and developing the service for its users. User Content is treated as confidential and is never disclosed to other users of the service, nor to third parties without prior written consent. The Results belong to the customer for normal business use, but never include Skene's methodology, code, algorithms or intermediary analysis. Product data itself stays in the customer's own Supabase or PostgreSQL database.
Reuse rights
The customer may freely reuse the Results for normal business operations without asking permission, including modifying them and creating derivative works. Two limits apply: the Results may not be commercialised as such, used to build a competing service, or reverse engineered to infer how Skene works. The analysis file Skene writes, skene-context/journey.yaml, is committed straight into the customer's own repository and remains readable without Skene. On its own side, Skene may use the Results to provide and support the service, to develop the service and other analytical tools, and to aggregate and combine them with other data. Retaining a copy of User Content is permitted for as long as the customer uses the service.
Data retention & training
Hosting summary
The privacy policy states that the servers and data related to the services are mainly hosted within the European Union. Where personal data is exceptionally transferred outside the EU or EEA, the company relies either on a European Commission adequacy decision or on appropriate safeguards such as standard data protection clauses. The jurisdiction is Finnish: the controller is Skene Technologies Oy, the terms are governed by Finnish law, and the competent supervisory authority named is the Finnish Data Protection Ombudsman. No individual hosting country is named on the site. The technical path is disclosed instead, service by service: Vercel hosts the dashboard and API, Supabase provides authentication, Postgres and storage, OpenRouter gateways the managed model analysis to OpenAI and Google, Resend sends transactional email, Stripe handles billing, and Upstash optionally buffers event ingestion. Several of these are United States companies, which sits in tension with the mainly-EU statement. Running the open source tool against a local model provider keeps everything on the user's own machine.
Things to keep in mind
Risks and trade-offs to weigh before adopting Skene.
- The privacy policy declares processing personal data to enhance, train and optimise AI models under legitimate interest, and no product-level opt-out from that training is documented anywhere on the site
- No Data Processing Agreement is published or offered, so anyone processing regulated personal data through the tool should obtain one contractually before starting
- Analysis prompts can carry repository, schema and journey context, and travel through OpenRouter to OpenAI and Google, so proprietary code context leaves your perimeter in the managed setup
- The pull request review is a model judging a diff against a policy; the vendor states it can be wrong and can contradict the deterministic gate, so treating its verdict as authoritative would be a mistake
- Trusting an automated tracking check can breed complacency: the tool cannot add an event nobody ever planned, and a green check means the plan was met, not that the plan was right
- Hosting is described as mainly within the European Union, but several named subprocessors are United States companies, and Vercel sees request traffic, session cookies and API payloads
- Contractual liability is capped at what you paid over the preceding six months, the service is provided as is with no warranties, and the vendor may change fees at its sole discretion with prior notice
Setup & Integrations
Technical difficulty
Low to moderate, and deliberately so. Cloud setup is three steps with no package to install: sign in with a magic link, install the GitHub App on the repositories you choose, then switch on pull request reviews. Nothing enters your dependencies or lockfile. Connecting Supabase adds an OAuth flow that only the workspace owner can complete. The command-line route is a single command through uvx with no prior installation. Expect to need repository access, a working understanding of your PostgreSQL schema, and possibly an organisation administrator's approval to install the app.
Deployment
Integrations
Behind Skene
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
Is there a free plan, and what does it include?
How much does the paid version cost?
Which code repositories does Skene support?
Does Skene replace PostHog or Mixpanel?
What can the Supabase connection actually see?
What permissions does the GitHub App request?
Can I use Skene without creating an account?
Is Skene SOC 2 certified?
Which third parties process data on Skene's behalf?
What happens when the monthly usage allowance runs out?
Should you pick Skene?
Skene is a deliberately narrow product, and that is its strength. It does not do analytics; it makes sure the tracking feeding your analytics is still there after the last release. The framing the vendor uses is persuasive: tests run in CI, code review comes from someone who did not write the diff, and tracking is the one form of correctness still left to the author of the change. Closing that gap with a deterministic check and a pull request comment is a sensible answer to a real and expensive problem, since lost event history cannot be recreated once the gap is found.
The engineering posture is unusually careful for a company this young. Nothing is pushed to your default branch, nothing enters your lockfile, the database connection starts read-only and cannot preview rows, ten subprocessors are named with what each one sees, and the command-line half is MIT-licensed and survives the vendor disappearing. The free audit costs nothing and does not expire, so evaluation is genuinely low risk.
The reservations are equally concrete. GitHub is the only supported repository host and PostHog and Mixpanel the only readable analytics sources, so a mismatched stack rules the tool out immediately. There is no SOC 2 certificate and no published Data Processing Agreement, which will stall some procurement processes. The privacy policy declares training and optimising AI models on personal data under legitimate interest without documenting any way to opt out, which deserves a direct question before signing. And this is a five-person company whose legal entity was registered in August 2025 and which is running on a pre-seed round.
For a product-led team on the right stack, arguing about whether behaviour changed or the tracking did, Skene is worth the free audit.
- Choosing a selection results in a full page refresh.
- Opens in a new window.